Correctly changing precision in DLRM

facebookresearch / dlrm

An implementation of a deep learning recommendation model (DLRM)

MIT License

3.71k stars 825 forks source link

Correctly changing precision in DLRM #376

Open rishucoding opened 6 months ago

rishucoding commented 6 months ago

Hello,

I would like to change the precision used in DLRM inference. I observed that the default datatype used in the original code is "float32", in the create_emb() method. Is changing the dtype to any of the supported values enough to correctly change the precision?

I believe that the generated embedding table would be directly used in the apply_emb() function.

Thanks for the help. Rishabh Jain

hrwleo commented 6 months ago

已收到邮件阿里阿豆故咋一马斯！如未及时回复，请致电15868848097 QQ：812737452